BACKGROUND OF THE INVENTION1. Field of the Invention
This invention relates to isolated dc/dc converters. In particular, this invention relates to low output voltage, high output current, isolated dc/dc converters that has multiple rectifier stages connected in parallel.
2. Discussion of the Related Art
In a high-power application, by connecting several substantially identical converter power stages in a parallel configuration to share the total power processed, one can often achieve a desired output power using smaller, lower-rated magnetic and semiconductor components. With several power stages connected in parallel, the power losses and thermal stresses on the magnetic and semiconductor components are distributed among the parallel power stages, thus improving conversion efficiency and eliminating “hot spots”. In addition, because lower-power, faster semiconductor switches can be used to implement the parallel power stages, the parallel power stages may be operated at a higher switching frequency than that of a corresponding single high-power stage. Consequently, the parallel configuration reduces the required sizes of the magnetic components and increases conversion power density. In addition, because the parallel power stages can be operated at a higher switching frequency, this approach can be used to optimize the transient response of a power supply.
FIG. 1shows converter100 with two forward-converter power stages101 and102 connected in parallel. Generally, a power supply with parallel power stages requires more power stage and control circuit components. However, if the parallel converters share the same output filter, the number of power stage components can be reduced, such as illustrated byconverter200 of FIG.2. Similarly, if transformer secondary windings are provided directly in parallel, required power stage components can also be reduced, such as illustrated byconverter300 of FIG.3.Converters200 and300 of FIGS. 2 and 3 are discussed in “Analysis, Design, and Evaluation of Forward Converter with Distributed Magnetics—Interleaving and Transformer Paralleling,” (“Zhang”) by M. T. Zhang, M. M. Jovanovic and F. C. Lee, published inIEEE Applied Power Electronics Conf. (APEC)Proc., pp. 315-321, 1995.
Regardless of the approach used in connecting power stages in parallel, ensuring that an acceptable load current (hence, power) is shared among the parallel modules is the main design challenge of such an approach. In fact, without an acceptable current-sharing mechanism, the load current can be unevenly distributed among the parallel modules. As a result, the modules that carry higher currents are electrically and thermally stressed more than the other modules, thus reducing the reliability of the power supply. Moreover, when the current of a parallel module exceeds its current limit, such as may occur when the converter current is unevenly distributed, the entire power supply may need to be shut off. Therefore, many current-sharing techniques of different complexities and performance are developed to ensure a relatively even current distribution among parallel modules. A discussion of some of these techniques is found in “A Classification and Evaluation of Paralleling Methods for Power Supply Modules,” by S. Luo, Z. Ye, R. L. Lin, and F. C. Lee, published inIEEE Power Electronics Specialists' Conf. Rec., pp. 901-908, 1999. For example, relatively even current sharing inconverters100 and200 in FIGS. 1 and 2 can be achieved by equalizing the peak values of primary currents in the modules. Furthermore, the performance ofconverter100 and200 of FIGS. 1 and 2 can be further improved by interleaving (i.e., operating the primary switches in each converter with 180° phase shift). Generally, as discussed by Zhang above, interleaving provides some input current and output current ripple cancellation, thus reducing the size of the input and output filters.
Referring to FIG. 3, steady-state current sharing amongparallel transformers301 and302 ofconverter300 is determined by the winding resistances oftransformers301 and302. Because winding resistance is usually comparable with the layout resistance, the current sharing performance of parallel transformers is sensitive to circuit layout. Sensitivity to layout resistance can be reduced by including a rectifier in the secondary side of each transformer, such as shown inconverter400 of FIG.4. Inconverter400, current sharing is determined by the on-resistances ofrectifiers401 and402, as a rectifier's resistance is usually larger than that of a printed circuit board (PCB) trace resistance. However, because the on-resistance of silicon rectifiers has a negative temperature coefficient (i.e., the rectifier's resistance decreases as the temperature of the rectifier increases), a current runaway condition may exist. In a runaway condition, all the secondary current flows through one of the rectifiers and the associated transformer secondary windings. The runaway condition inconverter400 can be avoided if low on-resistance MOSFETs (which have positive on-resistance temperature coefficients) are used instead of the diode rectifiers, as it is routinely done in low-voltage high-current applications.
In a low output voltage (e.g., below 3.3 V), high output current (e.g., above 50 A) application that requires transformer isolation, secondary-side conduction loss dominates total loss and limits conversion efficiency. Therefore, to increase conversion efficiency, rectification and transformer winding losses must be reduced. Rectification loss can be reduced, for example, by replacing Schottky rectifiers with MOSFET synchronous rectifiers. Reduction of transformer winding loss can be achieved by reducing winding resistance and the root-mean-square (rms) current in the winding, respectively, by properly selecting the winding geometry and transformer structure, and by employing a current-doubler topology. These techniques are discussed for example in “Design and Performance Evaluation of Low-Voltage/High-Current Dc/Dc On-Board Modules,” (“Panov”) by Y. Panov, M. M. Jovanovic, published inIEEE Applied Power Electronics Conf. (APEC)Proc., pp. 545-552, 1999, and in “The Performance of the Current Doubler Rectifier with Synchronous Rectification,” by L. Balogh, published inHigh Frequency Power Conversion Conf. Proc., pp. 216-225, 1995.
FIG. 5 shows an example of a 1.45-volt, 70-ampere dc/dc converter500 that employs a current-doubler topology implemented with synchronous rectifiers. (Converter500 is discussed in the Panov reference mentioned above). Inconverter500,synchronous rectifier501 and502 are each implemented by connecting three low on-resistance MOSFETs in parallel. The technique used inconverter500, however, cannot be extended to higher current levels by simply adding more synchronous rectifier MOSFETs, because the incremental reduction in conduction losses is less than the incremental increase of switching losses due to charging and discharging of MOSFETs' relatively large intrinsic terminal capacitances. If the switching frequency were not reduced, conversion efficiency would be reduced. However, reduction of switching frequency requires an undesirable increase in the sizes of magnetic components. In addition, the packaging of a large number of paralleled synchronous rectifiers is also difficult.
The output current ofconverter500 of FIG. 5 can be increased without efficiency degradation by connecting in parallel two or more power stages, as illustrated inconverter600 of FIG.6. However,converter600 requires significantly more power-stage and control circuit components to achieve even current (hence, power) sharing among the parallel modules. The additional components increase both the size and the cost of the converter.
SUMMARY OF THE INVENTIONAccording to the present invention, a parallel technique, which substantially reduces the number of power-stage and control-circuit components in an isolated dc/dc converter with a current-doubler rectifier and provides automatic current sharing is described. Using a common primary side inverter, and by providing in parallel only the secondary-side current-doubler rectifiers that are driven through separate isolation transformers, component count reduction is achieved. Current sharing among the parallel rectifier stages is achieved by connecting the primary windings of the transformers in series, thus forcing the same current through the transformers' secondary windings and the rectifiers. Additional component count reduction is achieved using integrated magnetic components. The technique of the present invention can be extended to an arbitrary number of rectifier stages, as well as to any rectifier topology.
The present invention is better understood upon consideration of the following detailed description and the accompanying drawings.
BRIEF DESCRIPTION OF THE DRAWINGSFIG. 1 showsprior art converter100 having two forward converter stages101 and102 connected in parallel.
FIG. 2 showsprior art converter200 having two forward converter stages201 and202 connected in parallel and sharing a common output filter.
FIG. 3 shows priorart forward converter300 havingtransformers301 and302 connected in parallel upstream torectifier303.
FIG. 4 shows priorart forward converter400 havingtransformers403 and404 connected in parallel downstream fromrectifiers401 and402.
FIG. 5 shows prior art half-bridge converter500 having a current-doubler output stage implemented bysynchronous rectifiers501 and502.
FIG. 6 shows prior art half-bridge converters601 and602, each having a current-doubler rectifier, connected in parallel.
FIG. 7 shows, schematically, dc/dc converter700, having an arbitrary number N of parallel rectifier stages, according to one embodiment of the invention.
FIG. 8 shows key waveforms ofconverter700 of FIG. 7, including (a) output voltage Vinvofinverter701; (b) secondary voltage Vsi, representative of a secondary voltage at one of transformers709-1 to709-N; (c) voltage VL1i, representative of a voltage across one of filter inductors703-1 to703-N; (d) voltage VL2i, representative of a voltage across one of filter inductors704-1 to704-N; (e) primary current iPRIM; (f) currents iL1iand iL2i, representative of the respective currents in one of filter inductors703-1 to703-N and in one of filter inductors704-1 to704-N; (g) current iD1i, representative of a current in one of rectifiers705-1 to705-N; (h) current iD2i, representative of a current in one of rectifiers706-1 to706-N.
FIG. 9 shows converter900 using magnetic coupling of output filters, in accordance with a second embodiment of the present invention.
FIG. 10 shows an implementation of converter900 of FIG. 9, using integrated magnetic components that has no magnetic coupling between filter inductors of the same rectifier stage.
FIG. 11 shows another implementation of converter900 of FIG. 9, using integrated magnetic components that has magnetic coupling between filter inductors of the same rectifier stage.
FIG. 12shows converter1200 withrectifiers1201 and1202, having integrated magnetic components on a single magnetic core (separation of core halves1203-1 and1203-2 is exaggerated for clarity).
FIG. 13 shows a model of the magnetic reluctance circuit ofconverter1200 of FIG.12.
FIG. 14 shows converter1300, which is an alternative implementation ofconverter1200 of FIG. 12, using opposite winding orientations to reduce the magnetic flux through center post1203-3. Note that the orientation of windings (dot positions) onlegs2 and3 are opposite to the orientation of the corresponding windings in FIG.12.
DETAILED DESCRIPTION OF THE INVENTIONIn the detailed description below, to facilitate illustration and correspondence between figures, like elements are provided like reference numerals.
FIG. 7 shows, schematically, dc/dc converter700 that has an arbitrary number N of parallel rectifier stages707-1 to707-N, according to one embodiment of the invention. Dc/dc converter700 usesinverter701 to convert the dc input signal into a bipolar high-frequency square-wave signal that is applied across the series connection of primary windings702-1 to702-N of transformers709-1 to709-N. Inverter701 can be implemented by virtually any converter topology, such as a forward converter, a two-switch forward converter, a half-bridge converter, or a full-bridge converter. As shown in FIG. 7,converter700 has secondary windings708-1 to708-N of transformers709-1 to709-N each coupled to a respective one of current-doubler rectifiers707-1 to707-N. Current-doubler rectifiers707-1 to707-N are connected in parallel at the output terminals ofconverter700. Of course, rectifiers705-1 to705-N and706-1 to706-N can be implemented by synchronous rectifiers, such as those discussed above with respect to FIG.5.
Because primary windings702-1 to702-N of transformers709-1 to709-N are connected in series, a common current iPRIMflows in all primary windings702-1 to702-N (assuming that the primary windings of transformers709-1 to709-N have identical magnetizing inductances). Consequently, if each pair of corresponding primary and secondary windings has the same turns ratio, secondary currents iSECin each of secondary windings708-1 to708-N are also the same, which ensures a perfect current (hence, power) sharing among rectifier stages707-1 to707-N. However, if the magnetizing inductances are different, secondary currents iSECwill also be different. Because the variation of magnetizing inductance can be easily kept within a narrow range, variations in magnetizing inductances do not significantly affect current sharing.
FIG. 8 shows representative key waveforms ofconverter700 of FIG.7. It should be noted that in FIG. 8 the symmetrical bipolar high-frequency voltage waveform at the output of the inverter implies that a symmetrical inverter topology (bridge-type topology) is assumed in the analysis that follows.
Ideally, when all components of rectifier stages707-1 to707-N are identical, the waveforms of signals in rectifier stages707-1 to707-N are identical. Thus, under ideal conditions, perfect current sharing is achieved, so that each rectifier stage carries 1/N of total load current iLOAD. Under ideal conditions, primary voltage Vpiacross each of primary windings702-1 to702-N is1/N input voltage V, or:
VP1=VP2= . . . =VPn=V/N
Initially, as shown in FIG. 8 between time t0to t1, voltage VINVof inverter701 (magnitude V) is applied equally across each of primary windings702-1 to702-N, thus inducing positive voltage Vsi=n*V/N across each of secondary windings708-1 to708-N, where n is the turns ratio across each corresponding pair of primary and secondary windings. (FIGS.8(a),8(b)) Consequently, rectifiers705-1 to705-N are each in an “off” state (FIG.8(g)), carrying no appreciable current. At the same time, a positive voltage VL1idevelops across each of inductors703-1 to703-N (FIG.8(c)), thus increasing inductor current iL1i(FIG.8(f)), which flows in the loop consisting of corresponding secondary windings708-1 to708-N, rectifier706-1 to706-N and filter capacitor710-1 to710-N. Because rectifiers706-1 to706-N conduct (FIG.8(h)), voltage VL2iacross inductors704-1 to704-N is negative and equals in magnitude to output voltage Vo(FIG.8(d)). As a result, inductor current iL2iin each of inductor704-1 to704-N is linearly decreasing (FIG.8(f)).
Between time t1and t2(i.e., time interval [t1, t2]), voltage VINVofinverter701 is zero (FIG.8(a)), inductor current iL1iin each of inductors703-1 to703-N, which was flowing during time interval [t0, t1] through corresponding secondary windings708-1 to708-N, continues to flow through rectifiers705-1 to705-N (FIGS.8(f) and8(g)). During time interval [t1, t2], voltage VL1ior VL2i(FIGS.8(c) and8(d)) across each inductor—i.e., any of inductors703-1 to703-N and704-1 to704-N—is negative and equal to output voltage Vo. Consequently, current iL1ior iL2iin each inductor is decreasing linearly at the same rate (FIG.8(f)).
During time intervals [t2, t3] and [t3and t4], the output voltage VINVofinverter701 is negative and zero, respectively. During these time intervals, the operations ofconverter700 are identical to those of time intervals [t0, t1] and time intervals [t1, t2], except that the roles of inductors703-1 to703-N and rectifiers705-1 to705-N are exchanged with those of inductors704-1 to704-N and rectifiers706-1 to706-N.
In rectifier stages707-1 to707-N, because voltage VL1iacross each of inductors703-1 to703-N is the same, inductors703-1 to703-N can be coupled, such as illustrated by coupledinductor901 of converter900 in FIG.9. (Similarly, because voltage VL2iacross each of inductors704-1 to704-N is the same, inductors704-1 to704-N can be coupled, such as also illustrated by coupledinductor902 of converter900 in FIG. 9) Using coupledinductors901 and902, the number of magnetic cores required to implement output filtering is reduced to two. Further reduction of the magnetic core count can be achieved by integrating coupledinductors901 and902 of FIG. 9 onto a single magnetic core, such as illustrated in FIG. 10 for converter1000 with two converter stages. Of course, the same concept can be extended to any number of rectifier stages. In the integrated magnetic implementation of converter1000 in FIG. 10, outer legs ofEE core1003 are gapped where the windings of coupledinductors901 and902 are placed. As shown in FIG. 10, the center leg ofEE core1003 has no gap and, therefore, has a much lower reluctance than the gapped outer legs. As a result, any flux generated in either of the outer legs is closed through the center leg (i.e., no coupling exists between opposite windings, so that there is no interaction between inductors703-1 and703-2 on one outer leg ofEE core1003 with inductors704-1 and704-2 on the other outer leg of EE core1003).
Alternatively, the magnetic integration of output filters can be also implemented by allowing a certain degree of coupling between filter inductors703-1 and703-2 wound on one leg of an EE core, and filter inductors704-1 and704-2 wound on the other leg of the EE core, as illustrated byEE core1101 ofconverter1100, shown in FIG.11. In FIG. 11, the coupling between inductors703-1,703-2 and inductors704-1 and704-2 wound on two outside legs ofEE core1101 is achieved by gapping the middle leg ofEE core1101. Due to an increased reluctance of the gapped middle leg ofEE core1101, relative toEE core1003 of FIG. 10, some flux that is generated in one outer leg ofEE core1101 is forced to flow in the other outer leg ofEE core1101, thus coupling all windings of inductors703-1,703-2.704-1 and704-2. When a proper amount of coupling is provided, the ripple in filter inductors703-1,703-2,704-1 and704-2 ofconverter1100 is less than the corresponding filter inductors in converter1000 of FIG. 10, thus improving converter performance.
Converter900 of FIG. 9 can also be implemented using a single magnetic core, such as illustrated byconverter1200 of FIG.12. Inconverter1200, 4-legged X-type magnetic core1203 is used. Note that, for illustrative purpose, core halves1203-1 and1203-2 are shown in FIG. 12 with an exaggerated separation. Actual separation between core halves1203-1 and1203-2 is typically a few millimeters, or less. In FIG. 12, core halves1203-1 and1203-2 implement coupled filter inductors703-1,703-2,704-1, and704-2 in the legs labeled “1” and “2”. Transformer windings702-1,702-2,708-1 and708-2 are implemented on the legs labeled “3” and “4”. To ensure correct operation ofconverter1200, magnetic core1203 is properly gapped, so that the fluxes created by the transformer windings are provided in the desired magnetic paths. To illustrate the gapping requirements, FIG. 13 shows reluctance circuit1300 that models the magnetic structure of core1203 of FIG.12.
Generally, in an implementation such asconverter1200 of FIG. 12, a magnetic coupling between the transformers and the filter inductors is not desired. Because filter inductors are intended to store energy,legs1 and2 of EE core1203 are gapped to create relatively large reluctances R1and R2, which are represented in FIG. 13 byrespective reluctances1303 and1306. In FIG. 13, inductors703-1 and703-2 inleg1 of EE core1203 are represented byvoltage sources1301 and1302, respectively. Similarly, inductors704-1 and704-2 inleg2 of core1203 are represented in FIG. 13 byvoltage sources1305 and1304. Because the transformers inconverter1200 are not intended to store energy,legs3 and4 need not be gapped. Reluctances inlegs3 and4 are represented in FIG. 13 byreluctances1312 and1309, respectively. However, without a gap, reluctances R3and R4are relatively small (i.e., reluctance R3and R4would each be comparable to reluctance Rcof non-gapped center post1203-3, which is represented in FIG. 13 by reluctance1313). Primary windings702-1 and702-2 are represented in FIG. 13 byvoltage sources1307 and1310, respectively. Similarly, secondary windings708-1 and708-2 are represented in FIG. 13 byvoltage sources1308 and1311. As a result of the relative reluctances of the transformers to those of the inductors, a part of fluxes Φ1and Φ2produced by inductor currents inlegs1 and2 of core1203 would flow throughlegs3 and4, in addition to the part of fluxes Φ1and Φ2flowing through center post1203-3. The amount of this flux coupling between the transformer legs and the inductor legs depends on the ratio of reluctance R3or reluctance R4to center-post reluctance Rc. To minimize this coupling, reluctances R3and R4should be made much larger than reluctance Rcby not having a gap in center post1203-3, and by introducing small gaps inlegs3 and4. The gaps inlegs3 and4 are generally much smaller than the gaps inlegs1 and2. In addition, when the air gaps are designed to achieve Rc<<R3=R4<<R1=R2, flux linkage betweenlegs3 and4 is also minimized (i.e., Φ3and Φ4corresponding to currents inlegs3 and4 are coupled to low-reluctance center post1203-3). As a result, currents in secondary windings708-1 and708-2 are each proportional to the respective current in primary windings702-1 and702-2 (i.e., the parallel current-doubler rectifiers707-1 and707-2 share load current ILOADequally). Otherwise, i.e., when fluxes Φ3and Φ4inlegs3 and4 are coupled, the currents in secondary windings708-1 and708-2 are not equal, even though the primary currents in702-1 and702-2 are the same, due to the internal impedance of each secondary circuit.
The flux in low-reluctance center post1203-3, which is shown in FIG. 13 as being equal to the sum of the fluxes of legs1-4, can be reduced by having opposite winding orientations in the windings of transformers inlegs3 and4, and in the filter-inductor legs1 and2. FIG. 14 shows such a configuration inconverter1400. (Note the difference between the dot positions of the windings in FIGS. 12 and 14.) With opposite winding orientations, both fluxes Φ3and Φ4and fluxes Φ1and Φ2flow in opposite directions through center post1203-3. As a result, the total flux Φcin un-gapped center post1203-3 is reduced, thus relieving reducing the area in center post1203-3 necessary to prevent saturation.
The integrated magnetic approach in FIGS. 10,11,12, and14 can be applied to any number of rectifier stages, although the integrated magnetic components in FIGS. 12 and 14 may require custom-designed magnetic cores when more than two parallel rectifier stages are present, because each additional rectifier stage requires an additional leg. For an even number of rectifier stages, the converter can be implemented with a number of x-type cores, using an x-core to integrate each pair of rectifiers, as illustrated byconverters1200 and1400 of FIGS. 12 and 14. Finally,converters700,900,1000,1100,1200, and1400 of FIGS. 7,9,10,11,12, and14 can be implemented using synchronous rectifiers, rather than diode rectifiers.
The current-sharing performance of each ofconverters700,900 and1000 was verified experimentally on a 200 kHz, 100 A/2.5 V prototype designed to operate from a 48-volt input. The prototype was implemented with a half-bridge inverter and two current-doubler rectifier stages. The measured full-load current-sharing performance and conversion efficiency are summarized in Table I.
| TABLE I |
|
| Measured current-sharing performance and |
| conversion efficiency of a 100-A/5-V prototype with |
| two paralleled rectifier stages |
| First rectifier | Second rectifier | |
| (i.e., rectifier | (i.e., rectifier | |
| 707-1) output | 707-2) output | |
| Implementation | current (A) | current (A) | Efficiency (%) |
|
| Non-coupled | 48.1 | 48.6 | 73.7 |
| inductors (e.g., |
| converter 700 of |
| FIG. 7) |
| Coupled | 48.7 | 47.8 | 73.7 |
| inductors (e.g., |
| converter 900 of |
| FIG. 9) |
| Integrated | 49.3 | 48.1 | 73.6 |
| Magnetics (e.g., |
| converter 1000 |
| of FIG. 10) |
|
The detailed description above is provided to illustrate specific embodiments of the present invention and is not intended to be limiting. Numerous variations and modifications within the scope of the present invention are possible. The present invention is set forth in the following claims.